Truace
Subscribe

Explainable machine learning framework for congestive heart failure detection using GLCM-based HRV features and Bayesian network analysis

BackgroundCongestive heart failure (CHF) is one of the main causes of cardiovascular morbidity and mortality. While conventional machine learning (ML) models demonstrate acceptable classification results, their operation is typically based on a black-box approach which does not allow to obtain an interpretable insight into the physiological mechanism of heart rate variability (HRV).ObjectiveThis paper suggests an approach which combines Gray-Level Co-occurrence Matrix (GLCM)-based texture analysis, Bayesian-opti…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1731· cite & verify
Explainable machine learning framework for congestive heart failure detection using GLCM-based HRV features and Bayesian network analysis
Shaher1 by Giardello. CC BY-SA 3.0 · https://creativecommons.org/licenses/by-sa/3.0

In brief

Researchers converted heart rate variability recordings from the PhysioNet database into two-dimensional representations and extracted Gray-Level Co-occurrence Matrix texture features. They selected the most discriminating features with the Kruskal-Wallis test and tuned machine learning models using Bayesian optimization.

The work matters because many machine learning classifiers for heart failure operate as black boxes, offering little insight into physiological mechanisms. Pairing a classifier with Bayesian network analysis is meant to make the diagnosis more interpretable, although the abstract describes only a database-based evaluation and not use in clinical practice.

Main points

  1. The best model was a Bayesian-optimized Quadratic Gaussian support vector machine.
  2. It reached 90.91% sensitivity and 94.44% specificity in classification.
  3. Bayesian network analysis ranked Dissimilarity, Cluster Prominence and Cluster Shade as key HRV texture biomarkers.
  4. The data came from the PhysioNet database, and features were chosen with the Kruskal-Wallis test.

The gain

A Bayesian-optimized Quadratic Gaussian SVM using texture features of heart rate variability separated CHF from non-CHF records with 93.10% accuracy and AUC of 0.9697, while Bayesian network analysis flagged the most influential features.

The rundown

The study addresses congestive heart failure, described as a main cause of cardiovascular morbidity and mortality. It argues that conventional machine learning models give acceptable classification but lack interpretable insight into heart rate variability physiology.

The method chains three steps: texture biomarker extraction from two-dimensional HRV representations, Bayesian-optimized model training, and Bayesian network inference. The network step used posterior probabilities and sensitivity analysis to examine nonlinear dependencies among the texture features.

The authors conclude that the explainable architecture could support early detection and personalized risk stratification of heart failure.

Sources

  1. Peer-reviewedTechnology and Health Care2026-10-08

The debate